Developing Measurement Models Using Bayesian Neural Networks

This paper presents the development and application of a Bayesian neural network (BNN) model to quantify the measurement uncertainty of non-contact articulating arms (NCAA) in measuring cylinder radii for criticality-based guard banding in quality control. Using real data, the study demonstrates the effectiveness of NCAAs, which produce point cloud data more quickly than traditional Coordinate Measuring Machines (CMMs), in alleviating manufacturing bottlenecks. Integrating this machine learning (ML) model into the manufacturing workflow significantly enhances process efficiency. Specifically, the model reduces reliance on CMMs in quality manufacturing processes, leading to faster operations and cost savings in quality control and reliability analysis of manufactured parts.